UROP Proceedings 2022-23

School of Engineering Department of Computer Science and Engineering 122 Knowledge Discovery over Database Supervisor: WONG, Raymond Chi Wing / CSE Student: LIU, Zihe / SENG Course: UROP1100, Summer Terrain is a class of algorithms which are used for generating, rendering, and analyzing three-dimensional terrain data. The aim of terrain algorithms is to provide the shortest path between two specific points on the model constructed by terrain data. Terrain Toolkit is a computer graphics library that provides a collection of algorithms and data structures for terrain modeling and rendering that enables the computing progress faster and more space-saving. The terrain toolkit can be divided into three parts: (1) surface simplification, (2) file format (3) shortest path computation. This paper investigates the application and improvement of surface simplification and shortest path finding procedures in the context of terrain toolkit. We adopt the surface simplification method proposed by Kaul et al. (2013) that ensures the shortest surface distance based on the simplified surface is within a bounded distance from the shortest surface distance based on the original surface. The shortest path computation is improved by integrating Xin and Wang's enhancement to Chen and Han's algorithm on the discrete geodesic problem and the Contraction Hierarchies (CH) algorithm. Besides, the article will show some preliminary idea for amelioration of toolkit, including QEM and A*algorithm. Knowledge Discovery over Database Supervisor: WONG, Raymond Chi Wing / CSE Student: MIN, Ruiming / MAEC Course: UROP1100, Spring Time series is a powerful tool for predicting future data based on past data. It can help us understand the patterns and trends in time-dependent data, such as sales, stock prices, weather, etc. However, time series analysis also faces some challenges, such as high computational complexity, limited data availability, easy overfit because of highly related, and uncertainty in forecasting. In this report, we will present some basic concepts of time series analysis and some methods to improve its performance. We will also demonstrate how to apply time series to some real-world problems and scenarios. Knowledge Discovery over Database Supervisor: WONG, Raymond Chi Wing / CSE Student: NIE, Weihao / SENG Course: UROP1000, Summer Recommendation systems are an excellent tool to filter information, which is widely used in e-commerce, social media, and other areas to optimize user experience. Since it was created, many scholars have researched it. Many datasets are proposed to help study the recommendation systems. Countless new methods are proposed to improve the performance of recommendation systems. However, I find that the improvement by updating methods on the existing datasets may be tiny now. Based on that, I found several datasets commonly used now in research. And try to propose new models through their raw datasets, which focus on the lost information of these datasets to improve the performance of the recommendation system.